基于《知网》的词语相似度计算研究

词语之间相似度的计算通常有基于统计和基于世界知识2种方法.中文词语相似度的计算,可以通过计算义原的相似度进而计算词语的相似度,在此基础之上深入分析《知网》的义原层次体系,提出了一种将义原在层次体系树上的深度和区域密度2个因素添加到义原相似度计算中的方法.通过实验发现,考虑义原在层次体系树上的深度和区域密度得到的结果比不考虑这2个因素得到结果更符合实际.

Research of Word Similarity Calculation Based on HowNet

There are two methods to calculate the similarity between words,which are based on statistics and world knowledge,respectively.HowNet could be used to calculate the similarity between Chinese words,which calculates the primitive similarity of HowNet first and further calculates the similarity between words.Based on the deep research on the primitive hierarchical system of HowNet,a method was proposed,which adds depth and area density of primitives on the hierarchical tree into the calculation of primitive similarity.The obtained experimental results find that the results obtained from this method are more practical than that ignoring the two factors.